Different measurement- and classification systems are used for quantifying PP, which is the most common skull deformity in infants. The majority of the gathered data on PP and the resulting classifications use two-dimensional measurement methods, as is the case with the classification by Moss and with that by Loveday et al.6,10. The use of this variety of methods is problematic from a clinical and scientific point of view because it complicates both the scientific discourse and the ability to reach an objective therapy decision.
Direct cephalometry—which is performed manually on the baby’s skull with a measuring tape and calliper—is common, although its reliability remains the subject of ongoing discussion1,8,24,25. Compared with this direct measurement of pure distances, stereophotogrammetry enables 3D data to be acquired and demonstrates a higher reliability as well as a very low artefact rate12,17,18,26,27,28. As a radiation-free procedure, stereophotogrammetry allows for repeated scans in the course of therapy and follow-up.
It is a reliable examination method and has good intra- as well as inter-examiner reliability, especially regarding the 3D parameters29.
However, it is still common practice to extract only 2D parameters from this 3D stereophotogrammetric set of data. This reduction of a three-dimensional object by using only two-dimensional parameters has a significant potential for error. In 2010, Lipira et al. demonstrated the extent to which even small deviations in the measurement plane can negatively affect the circumference and the measured distances14.
As shown in the present text, three-dimensional measurement methods have the potential advantage of more precisely representing the skull in its entirety19,20,30,31.
However, no appropriate 3D volume parameters or resulting indices have yet been described. Existing 3D parameters—such as ACAI or PCAI, which are used to categorise severities of PP—have turned out to be of limited suitability for clinical practice as they only represent the frontal or occipital volume quadrant and do not consider the entire three-dimensionality of the skull.
To the authors’ knowledge, none of the common, above-mentioned 2D or 3D classification have been further validated for clinical use thus far. One scope of our study was therefore to establish a suitable three-dimensional asymmetry index for clinical and scientific dialogue.
In addition, we generated a new classification using 3D stereophotogrammetric volume data for the first time by taking the three-dimensionality of the skull into full account. For this purpose, we performed a systematic statistical workup that investigated children both with and without PP.
Initial tests yielded highly significant distinctions between the patient group and the control group, thereby allowing for a clear differentiation based on the specific chosen parameters. To ensure comparability to the existing literature, we performed our analyses and created a classification system based on no, moderate, and severe PP, which is in line with Moss and Mortenson et al.6,11. Due to the widespread usage of the 30° diagonal difference described by Loveday et al.10, our measurements also originated from this gold standard. Separating the PP group by the 30° diagonal difference caused the ROC analysis that had been performed for this 2D diagonal parameter resulted in the highest possible value of 1, as shown in Figs. 6 and 7.
Prior to generating the presented data, we had performed a comprehensive pre-analysis in order to identify potential appropriate volume parameters. The two most promising volume parameters—|ln(Q4Q2/Q3Q1)| and |ln|Q3–Q4|—were depicted and evaluated in our study in order to facilitate a 3D-based classification of PP for the first time. In our ROC analysis parameter, |ln(Q4Q2/Q3Q1)| performed superior to the parameter of |Q3–Q4|, and we therefore agreed to use |ln(Q4Q2/Q3Q1)| for further statistical investigations. As |ln(Q4Q2/Q3Q1)| is a bulky label, we instead use the term 3D asymmetry index.
To differentiate between infants without and with moderate to severe PP, the volume parameter 3D asymmetry index with an AUC of 1 represented the maximum possible value. When differentiating between infants with moderate PP and infants with severe PP, the AUC was almost 1 (0.986). However, it still yielded the highest value, which means that no other 2D or 3D parameter had higher values in its respective group. The 30° diagonal difference of course scored an AUC of 1 (as shown above) because it defined the groups.
It should also be noted that the manual measurements with a calliper that were used to differentiate between moderate and severe PP also yielded very good results, with an AUC of 0.914. In differentiating between no PP and moderate + severe PP, these measurements even reached an AUC of 0.991. This finding highlights the strong results of this measurement method when used by experienced examiners, as has been reported in other studies11,25.
In their recently published report, Kato et al. evaluated the degree of severity in children with PP using 2D and 3D parameters. They could show that the severity assessment differs between 2 and 3D evaluation in one of six children. For 3D evaluation, they observed the anterior and posterior volume quadrants separately29. As shown above the comparison of both frontal volume quadrants (ACAI) resulted in considerably worse results than were found for all other parameters and proved to not be suitable. Clinical examination revealed that the extent of the frontal bossing did not correlate with the extent of the occipital flattening.
Although mainly the occipital region is affected in PP, the single juxtaposition of both occipital volume quadrants (PCAI) also yielded insufficient values to describe an existing asymmetry. This finding reveals that a single analysis of volume changes—either in the frontal or occipital quadrants—obviously does not solve the problem of reliably classifying PP.
Only a few studies have investigated ear shift. Kluba et al. demonstrated that only a weak correlation exists between the extent of the asymmetry and an ear shift32. In our own published data, we have also described ear shift as a weak parameter for classifying PP33. The conclusion of the presented analysis again stresses the notion that ear shift is not suitable as single differentiation criterion in PP. These results can be confirmed by clinical observations made by experienced examiners32.
Using the new volume parameter 3D asymmetry index, we established a new classification of PP using threshold determination. The 30° diagonal difference for differentiating between mild to moderate PP and moderate to severe PP—which is the actual gold standard in the existing literature—was maintained purposely for better comparability and easier applicability. After reclassifying the patients into three groups using the newly determined thresholds, we observed only minor shifts in the strength of each group when using the 3D asymmetry index as compared with the 2D parameter of 30° diagonal difference.
Our parameter 3D asymmetry index advances the development of the common classification systems used for PP. By describing no (< 0.0743), moderate (0.0743–0.2412), and severe (> 0.2412) degrees of PP, this index can be used as part of a new classification system that considers the entire three-dimensionality of the skull.
In accordance with existing studies, we included infants with comparable age and sex distribution. Although the number of included infants allowed for good statistical conformity, further validation with larger numbers in multicentre studies is needed.
As 30° diagonal difference is most commonly used to classify PP in present literature, we chose this parameter in the present study as well. Consequently, 30° diagonal difference reached the maximum possible AUC-value of 1. This means, that 3D asymmetry index can only be as good as 30° diagonal difference due to this methodology, which is a limitation in this study.
The comparison of relative values (ratios) in 3D parameters to absolute values in 2D parameters is a methodological limitation in this study. As mentioned above, we chose the 30° diagonal difference since it is the most commonly used parameter in literature as well as in daily clinical practice.
As a real three-dimensional parameter, a 3D asymmetry index might serve as a valuable basis for further studies that take clinical appearance and objective measurements into account, which has not yet been performed in the existing literature.